New approach for the sequential pattern mining of high-dimensional sequence databases

In this paper a new algorithm, the Top-Down mining of Sequential patterns (TD-Seq), for mining sequential patterns from high-dimensional stock sequence databases is presented. Existing algorithms are limited by efficiency problems in dealing with high-dimensional sequence databases. To address this problem, a two-phase mining method is proposed, in which a top-down transposition-based searching strategy as well as a new support counting method are exploited. Three pruning rules were also developed to reduce the search space further. Experiments conducted on actual databases demonstrate the improved performance of TD-Seq over existing algorithms.